基于K-means和TF-IDF的中文药名聚类分析  

Chinese drug name cluster analysis based on K-means and TF-IDF

作　　者：黄运高 王妍 邱武松[2] 向林泓 赵学良[2] 

机构地区：[1]重庆药品交易所股份有限公司,重庆404100 [2]中国科学院重庆绿色智能技术研究院高性能计算应用研究中心,重庆400714

出　　处：《计算机应用》2014年第A01期173-174,210,共3页journal of Computer Applications

基　　金：国家科技支撑计划项目(2012BAH19F01)

摘　　要：针对药名聚类中药物命名特殊性导致的命名准确率低的问题,提出了基于TF-IDF和K-means的药名聚类方法。药物命名具有一定的规律性且中西药名命名形式不同等特点,基于字词共现频率的方法难以取得较好的聚类效果,因此,使用TF-IDF方法计算药名相似的方法并采用K-means聚类算法进行药名的聚类。实验结果表明,TFIDF的聚类准确率高于TF的聚类方法,按字切分的聚类准确率高于分词后的聚类准确率,基于字和TF-IDF的聚类准确率最高且稳定,准确率达到96.77%。Because of the problem of low accuracy of Chinese name clustering, the method of durg name clustering based on TF-IDF （Term Frequency-Inverse Document Frequency） and K- means was proposed. As the durg name is with a certain regularity and western medicine is named in different forms, it＇s difficult to obtain better clustering results based on word co- occurrence frequency, so, TF-IDF method was used to identify similar drug names and K- means clustering algorithm was used for clustering drug names. Experimental results show that TF-IDF clusters drug names with high accuracy, the clustering of word-segmentation has higher accuracy than the clustering of participle. The clustering of words and TF-IDF has the higher accuracy and stablility, and its aeeuraey rate reaches 96.77%.

关 键 词：TF-IDF K-MEANS 中文药名聚类 药名分析 字词共现频率 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]